Signaling Generosity and Sickness Compensation Among the Ache

Theoretical Foundations of Food Sharing and Status Signaling

• The Evolutionary Paradox of High-Producer Food Sharing:

  • Daily resource pooling among foragers and horticulturists is widely interpreted as a risk-reduction mechanism ensuring adequate caloric intake across stochastic foraging returns.
  • Consistently high producers who donate substantial shares gain minimal risk-buffering from routine pooling because their baseline acquisition rates already preclude starvation on any given day.
  • Standard evolutionary models fail to account fully for individuals who consistently give away more than they receive, or average producers who share disproportionately high fractions of their yield.
  • The primary explanatory frameworks for biological resource transfers encompass:
    • Nepotism / Kin Selection: Resource provisioning structured by coefficients of relatedness ($r$) to maximize inclusive fitness (Hamilton, 1964).
    • Reciprocal Altruism: Conditional, tit-for-tat exchanges where short-term donation costs are offset by equivalent future repayments (Trivers, 1971; Rothstein and Pierroti, 1988).
    • Tolerated Theft / Scrounging: Transfers driven by asymmetric marginal values of food packages, wherein defending large resource packages against persistent demand becomes cost-prohibitive (Blurton Jones, 1987; Winterhalder, 1996).
  • Empirical studies across diverse ecologies demonstrate that these three models explain distinct components of food transfers but leave unexplained displays of generosity (Yanomamo: Hames, 1990, 2000; Gunwinggu: Altman, 1987; Ache in forest contexts: Kaplan and Hill, 1985; Hawkes, 1991; Ache on reservations: Gurven et al., submitted; Hadza: Hawkes et al., n.d.; Hiwi: Gurven et al., 2000).
  • Display sharing—transfers motivated by the quest for status or social display—demands alternative explanatory mechanisms.

• Reputation versus Status Displays:

  • Reputation is defined as an advertised cluster of personal traits that forms an intended or unintended mental representation of "self" among group members.
  • Different advertised traits (e.g., skill, reliability, generosity) yield distinct reputations, each potentially conferring high status through separate social pathways.
  • Existing models of status-driven food provisioning include:
    • Mating Effort / "Showing Off": Men target large, high-variance game packages that act as public goods to attract mating partners and gain social attention, operating under conditions where acquirers lack distributional control (Hawkes, 1991, 1992, 1993).
    • Offspring Survivorship Buffering: Successful hunters receive enhanced social attention, leading to preferential care and elevated survival for their children (Kaplan and Hill, 1985).
    • Political Power in Stratified Societies: High status functions as social power that secures privileged access to essential commodities during catastrophic subsistence crises, such as severe crop failures (Boone, 1998).
    • Costly-to-Replace Skill Insurance: Elite hunters among forager-horticulturists provision meat to establish social credit, ensuring support during prolonged debility (Sugiyama and Chacon, 2000).
  • Three fundamental theoretical questions govern investments in reputation:
    • Which observable behavioral displays establish a specific reputation, and how are these displays mapped into observers' beliefs?
    • What is the temporal stability of a reputation, and what schedule of ongoing behavioral investment is required to preserve it?
    • What is the temporal scale of fitness returns, and how do long-term compensatory payoffs trade off against immediate short-term benefits from alternative behaviors like resource hoarding?

• Food Sharing as an Honest Signal of Quality and Cooperative Intent:

  • Display sharing exploits pre-existing sensory and behavioral adaptations in recipients. This is analogous to courtship trembling in male water mites (Neumania papillator), which triggers female predatory ambush responses by mimicking the swimming vibrations of copepod prey near the water surface (Proctor, 1991).
  • Sharing serves as a dual signal communicating both underlying production capacity (quality) and willingness to engage in cooperative interactions (intent):
    • Signal of Intent: High-intensity sharing targeted consistently toward specific partners demonstrates reliable cooperative commitment, facilitating coalition formation, trade alliances, and dispute support.
    • Signal of Quality: High absolute volumes shared broadly across unpredictable recipient pools advertise superior foraging capability or stamina.
    • Honesty Enforcement: The signal remains evolutionarily honest because only individuals with consistently superior production can surrender high proportions of their yield over extended durations without starving themselves or their dependents.
  • Resource costliness modulates signal value. High-variance, calorie-dense foods like meat impose high foraging costs, making meat distribution a potent display currency among the Ache. Cultivated resources like manioc, available in large quantities, signal energetic investment in agricultural labor rather than acute physical skill.

The "It’s a Wonderful Life" (Bailey Effect) Model of Sickness Compensation

• The Problem of Periodic Incapacitation in Traditional Societies:

  • In the absence of physical food storage, sudden bouts of illness, accident, trauma, or venomous bites severely impede foraging and horticultural labor.
  • Morbidity and trauma rates across documented traditional populations are consistently high:
    • Ache (Arroyo Bandera reservation): Adults experience incapacitating illness on approximately $6.5\%$ of all days over a $3\text{-month}$ monitoring window.
    • Yora (Peru): Adult males experience sickness or traumatic injury on $8\%$ of days; the population-wide adult rate is $5.5\%$ (Sugiyama and Chacon, 2000).
    • Efe Pygmies (Ituri Forest, Democratic Republic of Congo): Men report physical ailments on $21\%$ of days; women report ailments on $22\%$ of days (Bailey, 1991). Traumatic injuries (plantar sores, thorns, hernias, and sprains) account for $63\%$ of these morbidities, equivalent to an Efe worker missing one full day of work per week.
  • Nutritional deprivation during acute trauma precipitates severe life-history penalties:
    • Prolonged muscular and physiological disability (Blackburn, 1977; Chang et al., 1998).
    • Stunted physical growth in recovering juveniles (Jackson, 1990; Kabir et al., 1998).
    • Elevated infectious morbidity and weakened bone health (Bonjour et al., 1997).
    • Delayed age at menarche (Maclure et al., 1991).
    • Subclinical ovulatory dysfunction and decreased female fecundity (Barr et al., 1994; Bringer et al., 1990).

• The Bailey Effect Mechanism:

  • Named after the protagonist George Bailey from Frank Capra’s 1946 film It’s a Wonderful Life, who prioritizes communal well-being over self-interest and is rescued from financial ruin by an outpouring of community donations.
  • The model posits that when temporary disability prevents food production, able-bodied group members selectively provision individuals who hold established reputations for high production and generous sharing.
  • This dynamic represents a long-term fitness exchange: short-term energetic losses from extensive food distribution function as an insurance premium against unpredictable, catastrophic periods of disability.

• Behavioral Typology in Reputation Space:

  • The model identifies four distinct behavioral strategies based on production volume and sharing fraction:
    • Philanthropic (George Bailey): Generates high production ($H$) and gives away a generous proportion ($s_g$).
    • Means-Well: Generates low production ($L$) and gives away a generous proportion ($s_g$).
    • Greedy (Mr. Potter): Generates high production ($H$) but retains most resources, giving away only a stingy proportion ($s_s$).
    • Ne’er-do-Well: Generates low production ($L$) and gives away only a stingy proportion ($s_s$).
  • Core Theoretical Predictions: Simple compensatory feeding directed toward generous individuals during down periods allows generosity to outcompete selfish hoarding across varied ecological conditions.
  • Mitigation of the Collective Action Problem: Unlike public-goods models of male hunting displays, this framework avoids the second-order collective action problem. If a donor’s marginal contribution accelerates a skilled producer’s recovery, the per-capita expected return in future food shares directly offsets the immediate cost of the aid.

• Formal Testable Predictions:

  • Prediction 1: Holding productive capacity constant, generous individuals will receive greater food assistance during incapacitation than stingy individuals (manifesting as more food per helper, aid from a larger number of helpers, or both).
  • Prediction 2: Holding generosity constant, individuals with higher baseline productivity will receive greater food assistance during incapacitation than low producers.

Ethnographic Setting and Methodological Design at Arroyo Bandera

• The Ache of Eastern Paraguay:

  • Historically full-time nomadic hunter-gatherers inhabiting a $20,000\,\text{km}^2$ range in the upper Jejui River watershed until peaceful pacification in the mid-1970s.
  • Despite transitioning to permanent settlements, modern Ache continue to allocate up to $33\%$ of their total time to extended mobile foraging excursions in subtropical forests (Hill and Tikuarangi, 1998).

• Study Population at Arroyo Bandera ("Flag Creek"):

  • Established in 1980 following an out-migration from Chupa Pou, located on the margin of a Protestant-administered Guarani reservation.
  • Positioned $1.5\,\text{km}$ from the western boundary of the Mbaracayu Reserve, over which the Ache retain exclusive foraging rights.
  • Demographics during the January–May 1998 field study (spanning the wet season into the early dry season):
    • $117$ total permanent residents.
    • $34$ adult males ($\ge 16\,\text{years}$ of age).
    • $25$ adult females ($\ge 16\,\text{years}$ of age).
    • $24$ dependent boys.
    • $34$ dependent girls.
    • Total residential composition: $24$ nuclear family households.
  • Settlement layout: Wooden-board houses built within the preceding $2\,\text{years}$, situated in a circular perimeter around a central soccer field.
  • Spatial distribution: Mean distance between any two residences was $108\,\text{m}$ (maximum inter-household distance of $242\,\text{m}$).

• Observational and Interview Sampling Protocols:

  • Food transfer events ($n = 380$) were monitored using three methodologies:
    • Random Time Blocks ($78\%$ of events): Systematically assigned 3-hour windows (07:00–10:00, 10:00–13:00, 13:00–16:00, or 16:00–19:00) tracking randomly clustered household subsets.
    • Focal Resource Follows ($10\%$ of events): Targeted tracking of large, irregular food packages brought from the forest or slaughter of large domestic livestock.
    • Systematic Interviews ($12\%$ of events): Structured recall capturing transfers occurring outside observation intervals.
  • Transfer classifications:
    • Source Distributions ($n = 228$): Initial distribution events where the original acquirer and donor belonged to the same nuclear family.
    • Subsequent Distributions ($n = 152$): Secondary transfer events where the donor and original acquirer belonged to different nuclear families.
  • Measurement procedures: Resource weights were measured directly using $10\,\text{kg}$ and $25\,\text{kg}$ Homs spring scales, or enumerated (e.g., individual stalks of sweet manioc) and converted to kilograms using standardized unit mass metrics.
  • Retrospective disability assessments: All accessible adults were interviewed regarding their most recent incapacitating illness or trauma ("last time so hurt that they had to stay in bed"). Interviews recorded duration of bed rest (days), medical symptoms, identities of all provisioning individuals, and specific food types provided.

• Variable Construction and Operational Definitions:

  • Compensatory Response Measures: (1) Total number of donors; (2) Number of donors outside the recipient’s nuclear family; (3) Percentage of donors outside the recipient’s nuclear family; (4) Total food items received; (5) Total meat items received.
  • Predictor Variables: Transformed into binary categories based on sample distributions:
    • Individual Generosity: Categorized by individual production shared with other households. The median cutoff was $87\%$ of individual yield given away ($\ge 87\%$ = "gave a lot"; $< 87\%$ = "gave little").
    • Individual Productivity: Ranked by aggregate production during random time blocks. The upper quartile ($25\%$) was classified as "high producers" because their yield was substantially higher than the remaining $75\%$ ("low producers").
    • Nuclear Family Generosity 1 (Fraction Given): Households donating more than the median proportion ($50\%$) of cumulative family production to other households.
    • Nuclear Family Generosity 2 (Total Network Reach): Households distributing to more than the median aggregate number of distinct households ($13$ families) across the entire sampling period.
    • Nuclear Family Generosity 3 (Distribution Breadth): Households sharing with more than the median average number of households per single distribution event ($2$ families).

Empirical Sickness Episodes and Assistance Patterns

• Epidemiological and Transfer Characteristics:

  • Of $50$ adults surveyed, $11$ reported zero recent episodes of bed-confining disability.
  • For the remaining $39$ adults reporting qualifying incapacitations, the median duration of disability was $3.5\,\text{days}$.
  • The median elapsed time between the disability episode and the survey interview was $30\,\text{days}$.
  • Diagnostic distribution of reported incapacitating conditions ($n = 39$):
    • Infectious and Internal Illness ($61\%$): Acute abdominal distress, coughing, febrile episodes, cephalalgia (headaches), and pulmonary discomfort.
    • Musculoskeletal Trauma and Wounds ($26\%$): Severe injuries to the ankle, arm, back, knee, shoulder, and hand.
    • Obstetric Events ($13\%$): Peripartum recovery and birthing confinement.
  • Nutritional composition of food aid ($113$ total items):
    • Carbohydrate crops: $67\%$ (predominantly cultivated reservation foods: sweet manioc, maize).
    • Meat portions: $29\%$ (wild game meat and domestic livestock).
    • Cultivated/wild fruits: $3\%$.
  • This aid composition diverges from baseline reservation intake, where meat represents only $10\%$ of total kilograms consumed and carbohydrates account for approximately $80\%$ of consumed biomass.

• Assistance Disparities by Gender and Age Classes:

  • Across the pooled sample ($N = 39$), a sick or injured adult Ache received an average of $3.26$ food items (including $0.74$ meat packages) from $2.44$ individuals, with $1.66$ of those givers coming from outside the patient's nuclear family ($53.3\%$).
  • Empirical distributions across demographic groups (Table 1):
    • Total Sample ($N = 39$): Donors = $2.44 \pm 0.30$ (range $1\text{--}9$); Outside Donors = $1.66 \pm 0.32$ (range $0\text{--}8$); Percent Outside Donors = $53.3\% \pm 6.8\%$ (range $0\text{--}100\%$); Food Items = $3.26 \pm 0.39$ (range $1\text{--}11$); Meat Items = $0.74 \pm 0.17$ (range $0\text{--}5$); Median Days Sick = $3.5$; Median Days Prior = $30$.
    • Males ($N = 22$): Donors = $2.82 \pm 0.48$; Outside Donors = $1.95 \pm 0.50$; Percent Outside Donors = $54.4\% \pm 9.0\%$; Food Items = $3.73 \pm 0.61$; Meat Items = $0.77 \pm 0.27$; Median Days Sick = $4$; Median Days Prior = $23$.
    • Females ($N = 17$): Donors = $1.94 \pm 0.29$; Outside Donors = $1.29 \pm 0.35$; Percent Outside Donors = $52.0\% \pm 10.6\%$; Food Items = $2.65 \pm 0.38$; Meat Items = $0.71 \pm 0.19$; Median Days Sick = $3$; Median Days Prior = $30$.
    • Younger Adults ($\le 35\,\text{years}$, $N = 21$): Donors = $2.71 \pm 0.50$; Outside Donors = $1.81 \pm 0.53$; Percent Outside Donors = $49.6\% \pm 9.2\%$; Food Items = $3.81 \pm 0.64$; Meat Items = $0.90 \pm 0.27$; Median Days Sick = $4$; Median Days Prior = $35$.
    • Older Adults ($> 35\,\text{years}$, $N = 18$): Donors = $2.11 \pm 0.29$; Outside Donors = $1.47 \pm 0.30$; Percent Outside Donors = $57.8\% \pm 10.2\%$; Food Items = $2.61 \pm 0.34$; Meat Items = $0.56 \pm 0.20$; Median Days Sick = $3$; Median Days Prior = $16$.
    • Younger Males ($\le 35\,\text{years}$, $N = 11$): Donors = $3.40 \pm 0.93$; Outside Donors = $2.20 \pm 0.99$; Percent Outside Donors = $47.6\% \pm 13.0\%$; Food Items = $4.90 \pm 1.19$; Meat Items = $1.20 \pm 0.49$; Median Days Sick = $4$; Median Days Prior = $40$.
    • Older Males ($> 35\,\text{years}$, $N = 6$): Donors = $2.33 \pm 0.40$; Outside Donors = $1.73 \pm 0.41$; Percent Outside Donors = $60.6\% \pm 12.7\%$; Food Items = $2.75 \pm 0.37$; Meat Items = $0.42 \pm 0.26$; Median Days Sick = $4$; Median Days Prior = $16$.
    • Younger Females ($\le 35\,\text{years}$, $N = 10$): Donors = $3.09 \pm 0.41$; Outside Donors = $1.45 \pm 0.51$; Percent Outside Donors = $51.5\% \pm 13.6\%$; Food Items = $2.82 \pm 0.44$; Meat Items = $0.64 \pm 0.24$; Median Days Sick = $4$; Median Days Prior = $30$.
    • Older Females ($> 35\,\text{years}$, $N = 12$): Donors = $1.67 \pm 0.33$; Outside Donors = $1.00 \pm 0.37$; Percent Outside Donors = $52.8\% \pm 18.5\%$; Food Items = $2.33 \pm 0.76$; Meat Items = $0.83 \pm 0.31$; Median Days Sick = $3$; Median Days Prior = $15$.
  • Wilcoxon rank-sum comparisons confirm that men received significantly more food items ($p < 0.10$) and aid from more total donors ($p < 0.10$) than women.
  • Younger men ($\le 35\,\text{years}$) received the highest levels of aid across all measures (food items, meat items, and total donors), whereas older women ($> 35\,\text{years}$) received the fewest donors and least assistance, matching relative social status hierarchies on Ache reservations.

Effects of Individual Sharing Depth versus Production on Aid Receipt

• Generosity as a Primary Driver of Sickness Compensation:

  • Individuals who surrendered above-median fractions of their personal production ($\ge 87\%$, $n = 19$) received more food items, more meat portions, and assistance from more non-nuclear family donors than individuals who kept higher fractions of their yield ($< 87\%$, $n = 20$).
  • High baseline productivity ($n = 10$ vs. $n = 29$) had minimal independent effect on assistance during illness or trauma.

• Test of the Four Behavioral Strategies (Table 2 & Figure 2):

  • Empirical allocation matrix of individual production:
    • Philanthropic ($n = 7$): Retained $4.8\% \pm 1.9\%$ of personal yield; donated an absolute mean of $68.8 \pm 19.4\,\text{kg}$ to other households.
    • Greedy ($n = 3$): Retained $15.2\% \pm 1.6\%$ of personal yield; donated an absolute mean of $42.8 \pm 17.7\,\text{kg}$ to other households.
    • Means-Well ($n = 12$): Retained $5.5\% \pm 1.3\%$ of personal yield; donated an absolute mean of $10.1 \pm 3.0\,\text{kg}$ to other households.
    • Ne’er-do-Well ($n = 17$): Retained $31.0\% \pm 5.1\%$ of personal yield; donated an absolute mean of $3.8 \pm 0.8\,\text{kg}$ to other households.
  • Production-Generosity Contingency: Out of $10$ high producers, $7$ ($70\%$) adopted the Philanthropic strategy. Out of $29$ low producers, only $12$ ($41\%$) adopted the Means-Well strategy.
  • Evaluation of Prediction 1: Strongly supported. Both Philanthropic and Means-Well individuals received significantly greater assistance across all five compensatory metrics compared to Greedy and Ne’er-do-Well individuals.
  • Evaluation of Prediction 2: Contradicted by empirical results. High baseline productivity failed to confer an advantage over low productivity when generosity was held constant:
    • Philanthropic individuals did not secure significantly more aid than Means-Well individuals.
    • Ne’er-do-Well individuals received higher assistance across every compensatory metric than Greedy individuals, despite Greedy donors transferring over ten times more food in absolute mass ($42.8\,\text{kg}$ vs. $3.8\,\text{kg}$, Table 2).
    • Means-Well donors also received significantly greater aid than Greedy donors across all metrics, despite transferring far less absolute food mass ($10.1\,\text{kg}$ vs. $42.8\,\text{kg}$, Table 2).
  • Relative Generosity and Fairness: Community members calibrate aid based on proportional sacrifice ($\frac{\text{Amount Donated}}{\text{Amount Produced}}$) rather than absolute volume transferred, favoring the honest intention to cooperate over sheer productive capability.

• Statistical Evaluation of Duration and Potential Confounders:

  • Duration of bed confinement correlates positively with four compensatory measures:
    • Total donors: Pearson's $r = 0.58$.
    • Donors outside the nuclear family: Pearson's $r = 0.55$.
    • Percent outside donors: Pearson's $r = 0.14$.
    • Food items received: Pearson's $r = 0.55$.
    • Meat items received: Pearson's $r = 0.65$.
  • These correlations are largely driven by two clinical outliers: one individual bedridden for $2\,\text{months}$ and another chronically disabled for $4\,\text{years}$. Both cases received food from $9$ donors and were classified as generous.
  • Multiple linear regressions incorporating both disability duration and the percentage of food donated to other families show attenuated beta coefficients for donation percentage:
    • Total donors beta shifts by $-0.43$.
    • Outside donors beta shifts by $-0.40$.
    • Percentage outside donors beta shifts by $-0.2\%$.
    • Total food items beta shifts by $-0.40$.
    • Total meat items beta shifts by $-0.25$.
  • Removing the two outliers attenuates mean differences but preserves positive directional trends and statistical significance in Wilcoxon rank-sum evaluations.
  • Reverse Causality Analysis: Generous individuals were confined for a median of $5\,\text{days}$, compared to $4\,\text{days}$ for stingy individuals. An examination of the percentage of food retained within the nuclear family before versus after sickness episodes revealed no significant shift, failing to support the hypothesis that individuals become generous only in anticipation of impending debility.

Nuclear Family Sharing Strategies: Breadth versus Consistency

• Household-Level Generosity Dimensions:

  • Analyses examined whether household sharing profiles impact aid received by an incapacitated member (Figure 3 and Table 3):
    • Proportional Household Sharing ($50\%$ median cutoff): Living in a household donating $> 50\%$ of its production produced modest gains in total food received, but significantly elevated the proportion of donors originating outside the immediate nuclear family ($p < 0.05$, Table 3).
    • Total Network Reach ($13$ families median cutoff): Donating to a higher cumulative number of households across the full season did not significantly increase aid received during disability episodes.
    • Distribution Breadth ($2$ families per event median cutoff): Routinely provisioning a higher average number of households per single distribution event significantly increased both total donors received ($p < 0.05$) and outside donors received ($p < 0.10$).

• Interaction of Depth and Breadth (Figure 4):

  • Cross-classifying households by proportional donation (depth) and average families reached per distribution (breadth) highlights the value of consistency:
    • High-depth, high-breadth households ($n = 9$) received the highest assistance across all metrics.
    • Low-depth, high-breadth households ($n = 13$) received substantially more donors and food items than households that shared with few families per distribution, regardless of whether those low-breadth families gave a large ($n = 10$) or small ($n = 7$) fraction of total food.
  • These patterns indicate that distributing food widely and consistently across each harvest or hunting event builds a more dependable insurance safety net than sporadic, high-volume transfers directed toward a small set of recipients.

• Summary of Statistically Significant Pairwise Group Comparisons (Table 3):

  • Total donors giving food: Males significantly greater than Females ($p < 0.10$); Younger Males ($\le 35$) significantly greater than Older Males ($> 35$) ($p < 0.10$) and Older Females ($> 35$) ($p < 0.05$); Generous individuals significantly greater than Stingy individuals ($p < 0.05$); Philanthropic significantly greater than Greedy ($p < 0.05$); Means-Well significantly greater than Ne’er-do-Well ($p < 0.05$) and Greedy ($p < 0.05$); Households giving to many families on average significantly greater than those giving to few on average ($p < 0.05$).
  • Outside nuclear family donors: Generous individuals significantly greater than Stingy individuals ($p < 0.10$); Philanthropic significantly greater than Greedy ($p < 0.05$); Means-Well significantly greater than Ne’er-do-Well ($p < 0.05$) and Greedy ($p < 0.05$); Households giving to many families on average significantly greater than those giving to few on average ($p < 0.10$).
  • Percent outside donors: Philanthropic significantly greater than Greedy ($p < 0.10$); Means-Well significantly greater than Greedy ($p < 0.10$); Low-proportion, high-breadth households significantly greater than low-proportion, low-breadth households ($p < 0.10$).
  • Total food items received: Males significantly greater than Females ($p < 0.10$); Younger Males ($\le 35$) significantly greater than Older Females ($> 35$) ($p < 0.05$); Generous individuals significantly greater than Stingy individuals ($p < 0.05$); Philanthropic significantly greater than Greedy ($p < 0.10$); Means-Well significantly greater than Ne’er-do-Well ($p < 0.05$) and Greedy ($p < 0.05$); High-proportion, high-breadth households significantly greater than high-proportion, low-breadth households ($p < 0.10$); Low-proportion, high-breadth households significantly greater than low-proportion, low-breadth households ($p < 0.05$).
  • Total meat items received: Generous individuals significantly greater than Stingy individuals ($p < 0.05$); Philanthropic significantly greater than Greedy ($p < 0.10$); Means-Well significantly greater than Ne’er-do-Well ($p < 0.05$) and Greedy ($p < 0.10$).

Comparative Evaluation of Competing Evolutionary Hypotheses

• Reputation Insurance versus the Showing-Off Hypothesis:

  • The showing-off model (Hawkes, 1991, 1992, 1993) treats large game as an indivisible public good where hunters lack distributional control, sharing indiscriminately due to tolerated theft to secure mating benefits or general social attention.
  • The Ache reservation data demonstrate that donors exercise active agency over transfer recipients and proportional retention.
  • The return currency in this system is concrete food aid during periods of physical disability, directly buffering health and survival rather than functioning solely as mating effort.

• Reputation Insurance versus the Skilled Hunter Hypothesis:

  • Sugiyama and Chacon (2000) posited that foraging communities provision the "best hunters" during illness to retain costly-to-replace foraging talent within the group.
  • The empirical Ache data show that high productivity alone fails to attract preferential care during debility. Greedy high producers were largely neglected relative to both Philanthropic high producers and Means-Well low producers.
  • These findings align with Sugiyama and Chacon’s secondary observation regarding the "loyal deputy": dependability and perceived cooperative fairness take precedence over raw foraging skill in eliciting community support.

• Extensive Food Sharing as Long-Term Risk-Averse Insurance:

  • Surrendering substantial fractions of hunted or harvested food imposes short-term energetic costs, resembling risk-prone or wasteful expenditure.
  • Over longer time horizons, routine food distribution functions as an insurance policy against the fitness risks of trauma and acute illness.
  • Condition-dependent strategic variation:
    • High producers possess greater energetic margin to absorb the short-term costs of food distribution, explaining why $70\%$ adopt generous strategies compared to only $41\%$ of low producers.
    • High producers who withhold resources (Greedy) may face high dependency loads within their own households or choose to sanction non-productive band members, trading long-term health insurance for immediate household nutritional security.

Mathematical Appendix: Formulation, Utility Dynamics, and Simulation Results

• Mathematical Model Structure:

  • Consider a population of size $N = 20$, divided equally into five individuals each of the four strategies:
    • Philanthropic ($nhh$): Production $H = 10$, Sharing proportion $s_g = 0.75$.
    • Means-Well ($nlh$): Production $L = 5$, Sharing proportion $s_g = 0.75$.
    • Greedy ($nhl$): Production $H = 10$, Sharing proportion $s_s = 0.25$.
    • Ne’er-do-Well ($nll$): Production $L = 5$, Sharing proportion $s_s = 0.25$.
  • In a baseline condition without disability compensation, every individual gives an equal share of their donated fraction to each of the remaining $N - 1$ individuals under tolerated theft.
  • In the absence of sickness, greedy strategists achieve the highest food accumulation, means-well perform worst, and philanthropic outcompetes ne’er-do-well whenever retained production satisfies:         $(1 - s_g)H > (1 - s_s)L$$(1 - 0.75)(10) = 2.5 > (1 - 0.25)(5) = 3.75 \quad (\text{condition not met in baseline parameterization})$

• Stochastic Disability and Compensation Architecture:

  • Disability risk is modeled by drawing an independent uniform random variable $p_{i,t} \sim U(1, 10)$ for each individual $i$ in each discrete time period $t$.
  • Incapacitation occurs if $p_{i,t} \le T$, where $1 \le T \le 10$. The per-period probability of disability is $\frac{T}{10}$.
  • Incapacitated individuals produce zero food during that time period ($0$).
  • During down periods, an injured individual who holds a reputation for generosity (Philanthropic or Means-Well) receives an added compensatory transfer $c$ from every able-bodied generous individual, and a partial compensation $\frac{c}{2}$ from every able-bodied greedy individual.
  • Ne’er-do-well individuals contribute $0$ in added compensation, and disabled stingy individuals (Greedy and Ne’er-do-Well) receive $0$ in added compensation.

• Per-Period Resource Accumulation Equations:

  • Let $n_{hh}, n_{hl}, n_{lh}, n_{ll}$ represent the counts of able-bodied individuals of each strategy, and let $d$ denote the total number of disabled generous individuals in period $t$.
  • When Healthy ($p_{i,t} > T$):
    • Philanthropic Strategist:             $\text{philanthropic}(t) = (1 - s_g)H - d c + \frac{s_g H (n_{hh} - 1)}{N - 1} + \frac{s_s H n_{hl}}{N - 1} + \frac{s_g L n_{lh}}{N - 1} + \frac{s_s L n_{ll}}{N - 1}$
    • Means-Well Strategist:             $\text{means-well}(t) = (1 - s_g)L - d c + \frac{s_g H n_{hh}}{N - 1} + \frac{s_s H n_{hl}}{N - 1} + \frac{s_g L (n_{lh} - 1)}{N - 1} + \frac{s_s L n_{ll}}{N - 1}$
    • Greedy Strategist:             $\text{greedy}(t) = (1 - s_s)H - d\left(\frac{c}{2}\right) + \frac{s_g H n_{hh}}{N - 1} + \frac{s_s H (n_{hl} - 1)}{N - 1} + \frac{s_g L n_{lh}}{N - 1} + \frac{s_s L n_{ll}}{N - 1}$
    • Ne’er-do-Well Strategist:             $\text{ne'er-do-well}(t) = (1 - s_s)L + \frac{s_g H n_{hh}}{N - 1} + \frac{s_s H n_{hl}}{N - 1} + \frac{s_g L n_{lh}}{N - 1} + \frac{s_s L (n_{ll} - 1)}{N - 1}$
  • When Disabled ($p_{i,t} \le T$):
    • Philanthropic Strategist:             $\text{philanthropic}(t) = d_{\text{comp}} + \frac{s_g H (n_{hh} - 1)}{N - 1} + \frac{s_s H n_{hl}}{N - 1} + \frac{s_g L n_{lh}}{N - 1} + \frac{s_s L n_{ll}}{N - 1}$
    • Means-Well Strategist:             $\text{means-well}(t) = d_{\text{comp}} + \frac{s_g H n_{hh}}{N - 1} + \frac{s_s H n_{hl}}{N - 1} + \frac{s_g L (n_{lh} - 1)}{N - 1} + \frac{s_s L n_{ll}}{N - 1}$
    • Greedy Strategist:             $\text{greedy}(t) = \frac{s_g H n_{hh}}{N - 1} + \frac{s_s H (n_{hl} - 1)}{N - 1} + \frac{s_g L n_{lh}}{N - 1} + \frac{s_s L n_{ll}}{N - 1}$
    • Ne’er-do-Well Strategist:             $\text{ne'er-do-well}(t) = \frac{s_g H n_{hh}}{N - 1} + \frac{s_s H n_{hl}}{N - 1} + \frac{s_g L n_{lh}}{N - 1} + \frac{s_s L (n_{ll} - 1)}{N - 1}$
    • Where $d_{\text{comp}} = c(n_{hh} + n_{lh}) + \left(\frac{c}{2}\right)n_{hl}$ represents total compensatory transfers received by an incapacitated generous individual from all able-bodied contributors.

• S-Shaped Utility Function and Collective Action Payoffs:

  • To capture nutritional and survival dynamics during clinical trauma, acquired food mass is transformed via an S-shaped utility function:         $V_i(t) = \frac{1}{1 + \left(\frac{1499}{100}\right) e^{-0.5 \times \text{amount}_i(t)}}$
  • This function imparts high marginal utility to food delivered during acute starvation or trauma relative to equivalent food consumed during healthy satiety.
  • An able-bodied donor will provide compensatory aid $c$ to $n_h$ or $n_l$ injured high or low producers if the discounted expectation of future shares upon recovery outweighs current losses:         $V\left(\frac{X}{N} - n_h c\right) + V\left[\frac{X + n_h s_g H}{N}\right] \ge 2 V\left(\frac{X}{N}\right)$$V\left(\frac{X}{N} - n_l c\right) + V\left[\frac{X + n_l s_g L}{N}\right] \ge 2 V\left(\frac{X}{N}\right)$         where $X$ is the aggregate production of all able-bodied group members.
  • When marginal assistance substantially increases recovery probability, per-capita returns are positive, eliminating the second-order collective action problem. Incurable or terminal conditions, however, will elicit no compensation because the patient cannot resume productive sharing.

• Simulation Results across Risk Regimes (Table A1):

  • Simulations were executed across $50$ time periods, repeated across $50$ iterations, in a population of $N = 20$ ($5$ agents per strategy). Parameters: $H = 10, L = 5, s_g = 0.75, s_s = 0.25$.
  • Payoffs are standardized to the highest-performing strategy ($1.00$) for each trial:
    • Trial 1 ($T = 3, c = 1$; $30\%$ morbidity risk, minimal compensation):
      • Philanthropic = $0.96$
      • Greedy = $0.83$
      • Means-Well = $1.00$
      • Ne’er-do-Well = $0.27$
    • Trial 2 ($T = 2, c = 2$; $20\%$ morbidity risk, moderate compensation):
      • Philanthropic = $1.00$
      • Greedy = $0.26$
      • Means-Well = $1.00$
      • Ne’er-do-Well = $0.09$
    • Trial 3 ($T = 2, c = 1$; $20\%$ morbidity risk, minimal compensation):
      • Philanthropic = $0.79$
      • Greedy = $1.00$
      • Means-Well = $0.82$
      • Ne’er-do-Well = $0.26$
    • Trial 4 ($T = 1, c = 2$; $10\%$ morbidity risk, moderate compensation):
      • Philanthropic = $0.97$
      • Greedy = $0.88$
      • Means-Well = $1.00$
      • Ne’er-do-Well = $0.22$
    • Trial 5 ($T = 1, c = 1$; $10\%$ morbidity risk, minimal compensation):
      • Philanthropic = $0.42$
      • Greedy = $1.00$
      • Means-Well = $0.53$
      • Ne’er-do-Well = $0.22$
  • Under high risk ($T = 3, 30\%$), a modest compensation ($c = 1$) maintains the evolutionary dominance of generous strategies.
  • At lower risk thresholds ($20\%$ and $10\%$), a compensation level of $c = 2$ ($20\%$ of a high producer's yield, $40\%$ of a low producer's yield) is required to ensure that the Philanthropic and Means-Well strategies outperform Greedy hoarding.